"Current" Games and replacing the benches.....

[BenSkywalker]

I have been noticing lately that the majority of people use very outdated titles as examples of "Current Games". Case in point, everyone is talking about how the GF3 is barely any faster in current games then the GF3 based on Quake3 benches. For those who are not aware, Quake3 came out in 1999, it is old now, not current. Evolva, OTOH, is showing the GF3 to be ~50% faster then the GeForce2 Ultra at the highest settings without FSAA. Evolva itself is now closing in on a year old, not exactly the newest title on the shelf for sure(picked it up at CompUSA for $9.99, had about ten different markdown tags on it)

This isn't just the GF3, I've seen many people basing advice with regards to performance on Quake3/UT benches as a guideline for current games.

Current games have upped the poly counts, increased texture sizes, they do use hardware T&L, they do use 64MB of RAM in some cases(Sacrifice refuses to run the highest texture quality settings without a 64MB board), they are adding Dot3. In short, they have advanced about a year and a half to two years beyond either of the standard issue benches that we see and what are used as represenative of "current" games. This month, not some far away time and place, Black&White is the number two selling game at EB and will be shipping in a few weeks. Tribes2 is shipping this month. Sacrifice and Giants have cleaned up in GOTY awards from many different sites and publications. All of them make UT and Quake3 look like what they are, rather dated games whose time, in terms of graphics, has come and gone.

I'm not saying that performance is irrelevant for people who mainly play those titles, as of late I've been playing Quake2 more then Quake3, but they are not and should not be held as a standard for current games.

As a replacement, I would suggest using something along the lines of MDK2 for OpenGL and Evolva for D3D. I'm not saying that these two are absolute(I'm open for suggestions), but they both have advanced far enough over the current standard issue benches that they are closer to what is a true current game.

"Bias" shouldn't be an issue anymore, with the exception of the Kyro(and rumors of a soon to be released Kyro2 are spreading) all of the gaming boards avaiable from active companies have hardware T&L and all of the features required to run the titles properly.

Maybe this is pointless rambling that noone will care about, but I think that it is a fair and true assesment of what is happening right now.

 

[Mykex]

Yea game benches have been pretty lame,then again so have game titles. For a few weeks I have been "dieing" to get a new game but I leave unimpressed and with empty hands each trip to purchase. I dont think you can really blame them for using older games since nothing has swept the masses in gaming for a while. Personaly I would like to see a bench or two using FS2000 like say bumpy ride into Boston adventure. Unfortunatly reviewers would be pulling there hair out screeming "I dont know I saw 56fps followed by 14 then back to 60 I think it hit 80fps once Im not sure". Anything that can smooth that game out when using full detail and a few different cloud layers w/rain-lightning landing at a busy AP gets my vote.....BTW,which game uses hardware T&L other than he he he Max Pane?

 

[Rellik]

Interesting.

I have been wondering about what games to benchmark as of late myself.
There are changes in benchmarking routines and they are closly linked with DirectX. When DX5 was out, Forsaken was THE benchmark. DX6 brought us Revolt(which is still used) and Drakan.

For D3D performance, I guess 3dMark 2001 will provide basic means to test performance. For true gaming I would recommend Mechwarrior 4.
It is DX8. I am not sure what features it supports but I will look into it.

One of the main reasons for testing Q3A in OpenGL was, IMHO, the sheer
number of different titles based on that engine that were good games:
-ST:Voyager
-F.A.K.K.
-Alice

Lithtech´s 2.5 engine supports some new effects(although i am not sure about special dx8 features) The engine powers No one lives forever.

My personal benchmark as of this writing is Q3 Team Arena Demo 127 for OpenGL, and UT with S3TC in OpenGL for a more cpu depended game.


Titles that are good candidates for being benchmarked:

- Giants
- Black and White
- Sacrifice
- Flashpoint
- Mechwarrior 4

If codemasters can fix their code(no pun intended) and properly implement cubic enviroment mapping and embm in Colin McRea rally 2,
then it would be a good bench for racing game performance.

 

[BenSkywalker]

Mykex

"BTW,which game uses hardware T&L other than he he he Max Pane?"

If you look at this thread, as I type this there is my reply and two other posts(in case someone else post while I'm typing this the only games mentioned that don't support hardware T&L are UT, Forsaken, Revolt, Drakan and possibly Flahspoint(not familiar with it). All the other do with a wide range of usefulness. From very mild(Quake3) to rather sizeable(Giants).

Rellik-

Out of your list for good candidates I have seen mention from the Sacrifice developers that they were considering a benchmark. I would imagine that it would be a bit different then actual in game performance as they would have to disable the variable geometry LOD, but it would be a good representation of the combined effect of graphics quality and FPS.

I'd love to see a Giants bench, although I did ask PlanetMoon directly about it and they said that it was not in their plans at that point in time(though maybe the GF3 build will give them a good reason to do it, they already have the FPS counter).

Mechwarrior4, I was under the impression that that was rather heavily CPU bound? I honestly haven't played the game yet so I'm speaking completely from what I have heard.

Black&White may do it, I have my copy on preorder as soon as I get it I will be sure to check for one

Not familiar with Flashpoint, do you have any linkage?

For OpenGL I have hopes that Tribes2 will include a benching utility. I see your point about the games that utilize the Quake3 engine, but even those, with the exception of Alice, are getting up there in age.

Another one that I have heard mentioned has been Nascar(Heat or 4, can never remember which one).

For synthetic benches we have both GLMark and 3DMark2K1 shortly, but most people view the synthetics with heavy scepticism unfortunately. GLMark is also rather lacking in some respects, no primitive tests like 3DMarkXX has given us over the years.

I think that one of the great reasons for people being unaware of what features are being utilized is four every six generations of graphics cards we see one update of benchmarks. A feature, such as T&L, that is in common use is believed not to be because none of the "latest" benchmarks use it. When Quake3/UT shipped we were right around the launch of the GeForce SDR, now we are looking at the GeForce3 using the same benches that became fashionable when the TNT2 Ultra was the hot ticket and told that "current games" aren't seeing any increase

Ohh yeah, just thought of a better OpenGL bench, SeriousSam. Engine looks extremely impressive.

The Lithtech engine doesn't seem too bad, but it also doesn't seem to be fill intensive. I have to say that after Kisssycho Circus I was full of doubt although NOLF has definately changed my mind on their abilities. Really an all around great looking game although it really isn't up to the level of Giants, Sacrifice, B&W etc.

 

[Adul]

Ben, you say tribes 2 is shipping this month. Where have your read or heard this?

 

[BenSkywalker]

Adul-

Linkage. They have, IME been very accurate as far as ship dates

 

[DavidJr]

Nascar Racing 4 requires a big and beefy system to run good on, I recommend that game for benching!!

 

[Adul]

Funny thing is I have not heard any news has to when tribes 2 is or will be done. They ahve been pretty much silent now. I hope that is still true, but I guess I will take a wait and see attitude. BTW, I not have seen any news of it going gold either.

 

[DeathKnightMaF]

Midtown Madness 2 supports hardware T&L. You can definately see what your card can do with detail turned to the max and drawing distance maxed out. Would be a good bencher for the high-end cards.

 

[BFG10K]

BenSkywalker:

I see were you're coming from, but the fact of the matter is dropping Quake 3 and UT as standard benchmarks doesn't make sense because they are still two of the most popular (and most widely played) games out there. Benchmarking new games that are not widely played is largely meaningless.

I'm willing to bet Unreal Tournament and Quake 3 will remain as standard benchmarks until they're superseeded by their own sucessors, namely Unreal 2 and Doom 3.

As a replacement, I would suggest using something along the lines of MDK2 for OpenGL and Evolva for D3D

In fact you'll find the vast majority of benchmarking sites are already using these games, in addition to Quake 3 and UT. Also MDK2 isn't that great of a choice given the Quake 3 engine is much more advanced and demanding, plus Quake 3 supports T&L as well.

 

[BFG10K]

Oh yeah, my vote for the current Direct 3D game would have to be Mercedes Beinz Truck Racing. That game will bring any system to it's knees.

 

[ndee]

BFG10K:
I think BenSkywalker meant, with replacing the benches, that the current card should be tested with current technology, and not with technology that is about 1.5-2 yrs old. Actually, it doesn't matter if the game is popular or not.

 

[BenSkywalker]

BFG-

"I see were you're coming from, but the fact of the matter is dropping Quake 3 and UT as standard benchmarks doesn't make sense because they are still two of the most popular (and most widely played) games out there."

I hear this argument a lot and while I can see that point, that tells you nothing about how current games are going to run. I gues part of this depends on what you are looking for in benchmarks, do you want to know about UT performance or D3D? There are a lot more people playing current D3D games as a whole then there are UT players(and even with UT, a great deal of those people are running either OpenGL(nV) or Glide(3dfx). Quake3 is a bit different as it does have several games based on it and does represent a large portion of OpenGL gaming performance. Even then, I think that Serious Sam would be a bit more represenattive of current games. Maybe Return to Castle Wolfenstein will include a bench utility that will make a good replacement for Quake3. Alice is already showing performance a decent bit lower then Quake3 now.

"Benchmarking new games that are not widely played is largely meaningless."

Not really. What are you going to play more with your new vid card, one game from two years ago or all the other current games combined? Back when the GF DDR was new the same two benches were in use that are being used today and I can tell you that I care a lot more about how Giants, Sacrifice, Tribes2, Serious Sam and B&W perform on it then either Quake3 or UT, even though for some time I was playing both of them heavily. It should be a complete given that any new card will handle every game out without problem at all, finding the most demanding/feature rich game will be more meaningful to users whether you are looking at a nVidia, ATi, Matrox or IT board. Unless you upgrade every generation how well a board performs with the absolute most recent and demanding titles should be, IMHO, more important then having a set standard.

"I'm willing to bet Unreal Tournament and Quake 3 will remain as standard benchmarks until they're superseeded by their own sucessors, namely Unreal 2 and Doom 3."

I wouldn't be shocked to see that myself, and it would be most unfortunate. This fall we could be looking at a Radeon2, GeForce4(or GeForce3 Pro/Ultra), and Kyro2 going head to head in UT and Quake3, really stressing all those new features As it stands now, we are seeing games with ~500%+ higher poly counts then UT, and a significant amount more then any Quake3 based game save the sole NV15 level. We are seeing Dot3 and other types of environment mapping used, we are seeing all the features that have been touted as being big step forwards in the latest games that make all the "hype" around the GF and Radeon a very useful reality, but how do they compare to each other in those titles?

My old DDR pushes the same FPS 1600x1200 32bit UHQ in Quake3 that it does running 640x480 16bit without hardware T&L in Giants, and in the standard UT benches I don't recall ever seeing numbers that low with my current system.

"In fact you'll find the vast majority of benchmarking sites are already using these games, in addition to Quake 3 and UT."

A few are that I have seen, but rarely if ever are they utilized as extensively as UT and Quake3. In the case of UT in particular it is truly sad as everyone has always known it isn't a good vid card benchmark in the first place. Do we really need to see another three pages of benches showing us every board performs nearly identical in nearly every setting available? UT isn't a very gfx card intensive game, never has been. Quake3 is much, much better in that aspect, but I wouldn't be too terribly interested in seeing scores on Quake2 or Quake1 on the current boards.

"Also MDK2 isn't that great of a choice given the Quake 3 engine is much more advanced and demanding, plus Quake 3 supports T&L as well."

Well, MDK2 uses a much better lighting engine then Quake3 and posts even better scaling IIRC. However, I'll gladly drop MDK2 from my list as I had forgotten completely about SeriousSam which is much better then either of them(have you checked out the test version yet?).

"Oh yeah, my vote for the current Direct 3D game would have to be Mercedes Beinz Truck Racing. That game will bring any system to it's knees."

Now that's what I'm talking about That is the type of benching I want to see. I could care less about how a GeForce3 or RadeonII or whatever else may come our way performs in two year old games, let's push them to their limits and utilize every lest bit of them we can instead of sticking to the old "fillrate is king" methodology. Show us how the boards perform in actual current games. Anyone who is going to drop a sizeable amount of cash on a bleeding edge board shouldn't pop it in to find out that the latest game they picked up doesn't run any faster. Conversely, as it stands now they could very well believe that that latest and greates won't give them any performance boost in that new title that is bringing their system to its' knees because Quake3 doesn't show any boost in the reviews they read.

DavidJr-

"Nascar Racing 4 requires a big and beefy system to run good on, I recommend that game for benching!!"

Does that one have a built in bench utility? Do you know what kind of advanced rendering features it may use? I want to check out all of the more recent titles with a decent bench utility that I can(not to mention I do need to pick up a newer racing title, been getting bored with the ones I have). MBTR and Nascar4 should take care of my speed fix for a bit

DeathKnightMaF

"Midtown Madness 2 supports hardware T&L. You can definately see what your card can do with detail turned to the max and drawing distance maxed out. Would be a good bencher for the high-end cards."

Does that have a bench utility included? The more the better for this

Adul

Not sure why nothing has been heard, I know the excitement over the title has waned recently. I was greatly anticipating it myself until I got a hold of Giants, now I will pick it up when it is available, but I certainly don't have the anticipation I did have.

ndee-

D@mn you said that a lot quicker then I did

 

[ndee]

BenSkywalker:
How long did you have to write such a long post?

 

[RobsTV]

Nascar 4 Racing has no specific and comparable test loop, but does have a fps counter. It does require a lot of horsepower. With all detail turned off, and at 640x480x16, and only 8 cars ahead, all else off or at the minimum, on the pace lap, from back of pack, driving alongside cars ahead, fps hover around 6 fps with a K6-2 392MHz and either V3 3000 or GF2 MX. CPU bottleneck there. K6-III @ 448MHz and GF2 MX, 640x480 at 16 bit "AND" at 32 bit, scores 20fps. Will test V3 with K6-III in a week.

With Duron @ 935 and GF2 GTS, at 1024x768x32 all at max settings, AA off, with 42 cars, hovers around 28fps.

 

[DavidJr]

For me with a 1.2ghz and geforce2 ultra with a ata 100 7200rpm hd, 256ram, with everything in the game maxed except draw ahead (set to 75%) with 42 cars ahead, 15 behind, 100% world detail, you name it its on, I get around 45fps at the back of the field, when I am by myself, i get around 95fps. But it takes a powerfull machine to run this game!! My old computer was a celery 400 with 128 ram and voodoo 3 3000 and it would get 5fps on it, it stunk!! No built in bench but you really need a good machine to run it!

 

[Rellik]

Ben: About Flashpoint:

Flaspoint 1985(working title: Operation Flashpoint) is a multiplayer
title like Team Fortress 2. It gives you the ability to take part in a
whole war campaign. from one man sniper mission to an air raid with
an A-10. Tanks, Helicopters, u name it. It doesn´t offer great effects like embm or cubic mapping I think, but it is a realistic engine with great skeleton animation. I think it will get quite a following if people would take the time to look aroud a bit more and realize that
there is more then Q3 or Counterstrike(although I´m not sure if I really like to see those cs campers in those servers.)

There are some videos on the net (3 or 4) and you can check out this LINK

I am currently working on a way to find out the criteria to a good systems check. The database will be filled with pc´s I get my hands
on, and I have to decide which benches to run to get a good picture of their performance. i just hope 3dmark 2001 isn´t so nvidia biased as the 2000 version....

 

[BFG10K]

BenSkywalker:

Well, MDK2 uses a much better lighting engine then Quake3 and posts even better scaling IIRC.

But Quake 3 requires more horsepower to run because it's a much more advanced engine. From what I've seen of MDK2, it's nothing more than a glorified version of Tomb Raider. I've seen a Celeron 300 running it with acceptable results, yet a Celeron 300 would absolutely die trying to run Quake 3.

but I wouldn't be too terribly interested in seeing scores on Quake2 or Quake1 on the current boards.

I would, as would a lot of others. Each new video card raises the bar of old games as well. Whenever I upgrade anything in my system, I'm interested in all of my 3D games, no matter how old they are (yes, that includes GLQuake). There are a lot of people with 1600 x 1200 (or higher) monitors, and each new generation of cards allows higher resolutions in older games.

For example I run Quake 2 at 1024 x 768 x 32 on my GF2 MX and I get ~95 FPS. I wouldn't run it any higher than that (if I could) because it'd be too slow. However if I upgraded to a GTS I could run the game at 1152 or 1280. If I had an GF2 Ultra I could run it at 1600 x 1200. The same applies to any older game. A faster card means more eye candy.

Also, as you mentioned UT is CPU bound so it really isn't a good video card test. A better one would be the original Unreal. From what I've seen, the engine is almost exactly the opposite of UT (ie it throws the burden almost exclusively on the video card). My statements about older games apply even more to Unreal. For a ~3 year old game, Unreal is still pretty demanding.

 

[BenSkywalker]

BFG-

"But Quake 3 requires more horsepower to run because it's a much more advanced engine. From what I've seen of MDK2, it's nothing more than a glorified version of Tomb Raider. I've seen a Celeron 300 running it with acceptable results, yet a Celeron 300 would absolutely die trying to run Quake 3."

Not really. It may not be the FPS that we like to see, but ~60FPS is what many people consider playable for Quake3. For MDK2, or any third person action game, lower FPS will seem smoother. As far as it being a glorified TombRaider, I guess it would be possible to argue the same about Giants But, I think SeriousSam would be much better, good old school Doom style gameplay(complete with tons of enemies coming at you from every direction) with the best OpenGL game engine I've seen to date.

"For example I run Quake 2 at 1024 x 768 x 32 on my GF2 MX and I get ~95 FPS. I wouldn't run it any higher than that (if I could) because it'd be too slow. However if I upgraded to a GTS I could run the game at 1152 or 1280. If I had an GF2 Ultra I could run it at 1600 x 1200. The same applies to any older game. A faster card means more eye candy."

I think you are overestimating how much of a strain Quake2 is putting on the latest graphics cards. My DDR gets ~60FPS 1600x1200 32bit color(mem bandwith is killing your FPS). GLQuake I can exceed my monitors refresh rate in every resolution without any problem. I would think that even a regular GTS would be in the ~70-80FPS range with an Ultra deep into the three figure FPS range in QUake2@1600x1200 32bit. There isn't an unplayable setting for either of those on my over a year old DDR(unless I crank up the res and FSAA).

"From what I've seen, the engine is almost exactly the opposite of UT (ie it throws the burden almost exclusively on the video card). My statements about older games apply even more to Unreal. For a ~3 year old game, Unreal is still pretty demanding."

They both use the same engine. The original Unreal is just as CPU bound as UT, it's just very poorly coded and hence moves much faster under Glide then D3D or OpenGL(although the last patch they released aided significantly). Demanding Unreal still is, unfortunately it is because of how poorly the graphics engine was coded, not because it is taxing the system with loads of eye candy.

Rellik

DLing the movies now(Dial-Up, will take a while). Looked it over and it does look impressive in terms of the player models at least, I'd imagine with the heavy AI required for this game it is going to require a heavy system for smooth play. I like hearing about skeletal animation being used, means it should have great hard T&L support, maybe we will see some selectable geometric LOD with a decent range.

"I am currently working on a way to find out the criteria to a good systems check. The database will be filled with pc´s I get my hands on, and I have to decide which benches to run to get a good picture of their performance."

Sounds interesting, are you doing that for a site?

"i just hope 3dmark 2001 isn´t so nvidia biased as the 2000 version...."

You can blame that on 3dfx and ATi(seriously). The problem for 3D2K was that they had to use 16bit as default(remember when it came out, 3dfx still didn't have a 32bit capable card) and ATi's horrible 16bit performance. Compare 32bit results at the highest resolutions and the V5, Radeon and GF2 are all pretty close. This time around, 32bit is the default so ATi will look much better compared to GF2 level hardware, unfortunately they are likely to look very poor stacked up against the GF3. One interesting point for this years version is that Max Payne will actually be out within months of the new bench, and by the looks of it I wouldn't be surprised to see several games based on the MP engine popping up.

ndee-

Well, about two hours, but there is a reason for that. Where I live we were getting hammered by a snow storm and I had to go out and help get a vehicle out of a snow bank. I started typing my response about an hour before you posted yours(BFG's post was fresh) and finished it about an hour after

 

[hans007]

If a game that is similar to say mdk2 in the future comes out, and it was slow with our current cards, we'd just buy new cards, when that game comes out right? most people buy a card, to play a game that is already out that they want to play. The only reason to bench a game that most people dont play such as merceded benz truck racing, which i'm sure not too many people play regularly as opposed to a game like counterstrike, is to find out how future games that people will play say a counterstrike 2 on current cards. That is just dumb, because benchmarking now for something that doesnt exist is illogical, since people dont buy cards to play games that aren't out. If people just bought cards when they needed them for the next game, everything would work out, since if you buy an EXTRA powerful card now that you wont put to use with current games (such as the gf3) then you are just letting its extra features and extra price, be useless until a future game comes out. And by then this card will be cheaper, and you'll just have wasted money and be an idiot. Sure some headroom in the card you buy is good, maybe it can last two generations of games, but in that case we should bench them with both generations of games, i.e. mercedes benz, and quake3

 

[BenSkywalker]

hans007-

"If a game that is similar to say mdk2 in the future comes out, and it was slow with our current cards, we'd just buy new cards, when that game comes out right?"

No, absolutely not. Why are you running a 1.1GHZ and Radeon DDR, both overclocked from default specs?

"most people buy a card, to play a game that is already out that they want to play."

No, they don't. Most people buy a board to improve performance for lots of games(if they are a gamer anyway), not just one. If you would drop ~$100 or more, sometimes significantly more, on hardware to play one game then it had better be one he!l of a game.

"The only reason to bench a game that most people dont play such as merceded benz truck racing, which i'm sure not too many people play regularly as opposed to a game like counterstrike, is to find out how future games that people will play say a counterstrike 2 on current cards."

No. MBTR utilizes features and reflects performance common amongst current games, not just MBTR. If you only buy one game once every few years, like the gap between CS and CS2 will be, then why would you upgrade at any point between them? Giants, Sacrifice and MDK2 were all mutliple GOTY nominees and award winners, lots and lots of people are playing these games. Counterstrike tells you absolutely nothing about how they would perform, neither would UT or Quake3. MBTR, OTOH, gives a much better indicator.

"That is just dumb, because benchmarking now for something that doesnt exist is illogical, since people dont buy cards to play games that aren't out."

That's BS, at least as a blanket statement. I picked my DDR with every intention of it lasting me a long time and offering considerably better performance long term then anything out at the time for future games. I was correct. Even looking at the V5 5500 which came out six months later and was a much "faster"(in 1999 games), at the settings I play at with the games I am currently playing the DDR is quite a bit faster not to mention offering superior visual effects. Back in 1999 when I decided to purchase one that was looking forward to games that weren't even out yet and features that were not being used, today those features are being used in the majority of new 3D games. This isn't predicting some far off time, this is now. I'm not saying revert to 3DMark2K1 and GLMark as examples of what we will see in the future, I'm talking about using existing, already shipping games which are comparable to actual current games to bench instead of circa 1999 titles. The V3 is comparable to the GF2MX in UT benches and in quality, fire up Giants and the V3 looks and runs like complete sh!t next to one. 1999 vs 2001.

"If people just bought cards when they needed them for the next game, everything would work out, since if you buy an EXTRA powerful card now that you wont put to use with current games (such as the gf3) then you are just letting its extra features and extra price, be useless until a future game comes out."

How many times have you upgraded your video card in the last year? How much did it cost you total? Zero upgrades for me in that time frame, although fourteen months ago when I picked up my DDR it ran me $320. I bought the fastest card on the market bar none(I got the Herc which clocked at 130/301 vs 120/301), got a lot of useful features for myself(hardware T&L- visualization) and put myself in a position that I didn't need to upgrade for over a year without problems and still being able to run every game that came out at the highest settings(though not always the highest res). If noone buys cards with feature sets beyond what is absolutely bare minimum to run the latest game, then we will not progress.

Besides that, MBTR is an example of an already shipping game that uses features that most new 3D games are using. This isn't just about the GF3 either, your Radeon would look a lot better stacked up against the GF2 in Evolva then it does in UT

"Sure some headroom in the card you buy is good, maybe it can last two generations of games, but in that case we should bench them with both generations of games, i.e. mercedes benz, and quake3"

Last two generations so bench games that are one generation and three-four generations old???? That makes no sense at all. If you want a card to last you two generations into the future you shouldn't look to games from the past. You should look to the latest games that you can. If you want to know how a game is going to perform a year in the future, why in the he!l would you a bench a game from two years in the past???????:Q

I'm honestly very interested in hearing your thoughts on this, I have seen your line of thought before expressed by others(interestingly enough, seems to be the people who buy three different vid cards every year), and I have to say that it makes absolutely no sense to me at all. I can see the arguments put forth by others, but not one justifying benching two year old games to see how a board is going to last for another year.

 

[hans007]

i'm talking about the line of thought that benchmarking is supposed to help people who want to see what they will need to have now, to run games of several generations. I am not saying i am one of these people. I just buy stuff because i'm addicted to buying faster stuff. I mean really i used to have an 800 tbird, and a g400, and well the g400 blew ass in win2k counterstrike so i bought a radeon. The 1000@1100 has had no noticeable effect and well i just bought it. figure i overclock the radeon since i already have it and might as well do it. so sometimes i buy because i need speed and sometimes i just buy stuff.

Now for people who use reviews with benchmarks as something of an indicator of what they will do for a purchase, you should bench old games such as quake and new games such as mbtr. Because a new card may not have the same improvement effect on the older games as much as on the newer ones and well it could help your buying decision. If people are really concerned with the games of today and are hard core quake3 players (sure its an older game, but a ton of people still play it same with UT) they'd probably buy a geforce2 even if the radeon is better at mbtf (i wouldnt know since i havent really researched this) because it is more important for them that quake3 run better not the future games. This is for people who use benchmarks really to judge what they want to buy. the argument that we shouldn't bench a game if its really old would fit for like q1/q2 since people aren't reallyplaying that very much anymore same with the original unreal. I really think a consumer wants to buy the better overall card for their needs. And a combination of older games like quake3 and newer games and even unreleased games in the not too distant future are what will affect what that consumer needs. Thus the need to continue benching quake3. A card that is twice as fast (same a gf2 ultra vs a gf2) at some new game, might not show the same amount of difference in an older game or not enough of a difference to the consumer, and a consumer would like to know what those numbers are to justify paying more for say the ultra or gf3 or whatever. They cant weigh their options as well with less data.

me personally in the past year, i've had a v3 a geforce ddr 64mb that i traded for a voodoo5 a g400 for a while when i needed money and now a radeon. I upgraded from the v3 since i got the geforce as a free sample. my v3 is in my cousins computer and i sold the g400 for about what i paid for it since i only had it a month. SO really i've spent less than you and still have 2 cards that i've actually paid for. I probably wouldn't have gotten the geforce or consequently v5 because it cost too much. a v3 couldn't run counterstrike at 1024x768 but i used it primarily for UT at the time, just started playing CS after summer and i had the v3 last winter. you cant just settle in on one game that represents all opengl games or something like that for this reason. personally i dont play evolva or mbtr, i only play counterstrike, so if i saw a bunch of reviews with only mbtr and evolva benches it would be hard for me to make a decision on which card was better at counterstrike if say a radeon was better at mbtr but a geforce2 was better at evolva. if i played evolva then the choice would be easier.

I usually buy something new if something seems terribly slow at a setting i want to run and i have money. If i have a lot of money somehow i might just buy something new just because its new. kinda like how people buy new cars even though they dont really need a new one and their old car is fine for just drivng, because they can afford it, and they'd keep their old car if they couldn't afford it.

 

[BenSkywalker]

Hans-

"so sometimes i buy because i need speed and sometimes i just buy stuff."

Good reason, but that means you don't wait until a game won't run I take it?

"If people are really concerned with the games of today and are hard core quake3 players (sure its an older game, but a ton of people still play it same with UT) they'd probably buy a geforce2 even if the radeon is better at mbtf (i wouldnt know since i havent really researched this) because it is more important for them that quake3 run better not the future games."

Benchmarks, except on game specific sites, are not supposed to be about how well a given card performs on one particular game. More people are playing Couterstrike right now then Quake3(yes, I would bet on it), but when is the last time you have seen CS benched? The utility to bench HL works if you use the proper build(most sites are still using build 1.17 of Quake3 to bench). Why, if you are honestly looking for games performance that people play the most often, wouldn't you use the game that people play the most often?

"This is for people who use benchmarks really to judge what they want to buy. the argument that we shouldn't bench a game if its really old would fit for like q1/q2 since people aren't reallyplaying that very much anymore same with the original unreal."

Last few times I checked, Quake2 had more people playing then Quake3, sometimes by a more then 2 to 1 ratio. Even the original Quake wasn't doing to badly, although the original Unreal was completely dead. But even if that wasn't the case, every current gen board out right now plays Quake3 d@mn well, how long do we go on using it as a benchmark? CS is more popular and the engine has been benchable for a long time now.

"A card that is twice as fast (same a gf2 ultra vs a gf2) at some new game, might not show the same amount of difference in an older game or not enough of a difference to the consumer, and a consumer would like to know what those numbers are to justify paying more for say the ultra or gf3 or whatever. They cant weigh their options as well with less data."

How many people play Quake3, or any other singlar title, to care that much about it as a sole indicator of performance? That is what we are seeing now in reviews. I agree that if a site wants to include twenty five pages of benches in there(which you can check on these boards, I have been a supporter of) throw Quake3 and Quake2 in, he!l yeah. If, however, they are going to do what they currently do, rely on two to four games then make them all recent ones. Any upcoming board is going to run Quake3 incredibly well, that is a given.

"me personally in the past year, i've had a v3 a geforce ddr 64mb that i traded for a voodoo5 a g400 for a while when i needed money and now a radeon. I upgraded from the v3 since i got the geforce as a free sample. my v3 is in my cousins computer and i sold the g400 for about what i paid for it since i only had it a month. SO really i've spent less than you and still have 2 cards that i've actually paid for."

In the last year I have spent nothing at all Besides that though, you got a ~$300-$350 board for free, that is a bit of odd circumstances. If you had bought the board in the first place, and then followed through with all of your trading you would be out ~$400 possibly more. Not criticizing at all, just seems that people who don't think future performance for boards is important tend to be people like yourself, those that go through five different boards in a single year. If you do that every year, I can certainly understand why it doesn't matter to you. There was a thread over in GH where a person was aksing for advice on a board to pick up to last him a year or two, people were reccomending the V5 to him I'd like to dig up that post a year or two from now and see how many of those same people will still be using a V5. What reasons were given for making the reccomendation? In summary, "current games performance" based on the performance of games from two years ago. The major sites post benchmark scores and people take that as indicative of current game performance, even if it is nothing at all like how the boards stack up in actual current games.

Look no further then this thread. A Senior Member on this forum wasn't aware of any T&L games? The majority of new titles shipping use hardware T&L but where is the average user going to find that out? It isn't mentioned a single time in the Sacrifice manual IIRC, but every time I load the game I have to chose hardware or software T&L. What if I didn't have a hard T&L board? I wouldn't be any wiser to the fact that the game supported it. Similar situation with nearly every other T&L game that I have purchased, unless you have a T&L board, you aren't going to know that the game supports it and most of the time could be running much better using it.

Then there are the added features of boards like the GF2 or Radeon. Right now, there are many games available that look much, much better utilizing features found on these two boards but not on something like the V5, when is the last time you have seen a perspective buyer of one of these boards being warned of that fact? They aren't, and why not? "Current games"(1999) don't use those features according to the masses. This is a perception factor along with being an informative one, game companies don't want to tell you you can't get all the goodies in this title without either xxx or xxx video card, so who is supposed to? If benches need to be run and show that card xxx can't run this test because it doesn't support this feature, no matter who it is that has to fail, shouldn't the consumer be made aware that he can't see everything in that new game he purchased? This is why I think we should drop Quake3 and UT. Large textures and 32bit color were the "major" new features Quake3 brought to the table, both were available from everyone except 3dfx for a long time before it debuted. 3dfx is now dead, we can stop having to find benches that they can handle properly to appease the loyalists.

"you cant just settle in on one game that represents all opengl games or something like that for this reason"

But that is exactly what is happening right now and what's worse, it is a two year old represenative.

"personally i dont play evolva or mbtr, i only play counterstrike, so if i saw a bunch of reviews with only mbtr and evolva benches it would be hard for me to make a decision on which card was better at counterstrike if say a radeon was better at mbtr but a geforce2 was better at evolva. if i played evolva then the choice would be easier."

Neither Evlova nor MBTR would be good for CS performance(unless you run D3D), but SeriousSam would likely be fairly decent within a small range. CS is more CPU intensive then anything else and unless you run actual CS benches, not even Quake3 is going to be a great indicator. Overall OpenGL performance is what reviewers are supposed to be shooting for, instead we have a Sweeny/Carmack test set and not much else. WTF does that tell us about overall D3D and OpenGL performance?

"If i have a lot of money somehow i might just buy something new just because its new. kinda like how people buy new cars even though they dont really need a new one and their old car is fine for just drivng, because they can afford it, and they'd keep their old car if they couldn't afford it."

What about those who either can't afford or refuse to spend money upgrading constantly? People who upgrade vid cards less then once a year are a majority even amongst gamers, rich or scraping by. I have the money right now to order a GeForce3 from Japan and have it shipped here, I'm not going to though. To me it would be a waste of money until I'm quite certain it is going to be my best bet in the near future for an upgrade that will allow me to avoid upgrading again for another fifteen to sixteen months. I'll build a whole new system around a vid card(already have the DX8 compliant vid card platform running), but once I get it I keep it for quite a while.

I can see now where you are coming from, though I think that what you are looking for in benches is quite different then what there intended purpose is.

 

[BFG10K]

BenSkywalker:

Not really. It may not be the FPS that we like to see, but ~60FPS is what many people consider playable for Quake3.

I've already seen the benches but I've also seen how it runs in real life. 60 FPS in Quake in real life sucks. 60 FPS in MDK2 in real life is more than playable.

I think you are overestimating how much of a strain Quake2 is putting on the latest graphics cards.

Not at all. All I know is how well it runs currently on my system. I also know that if I upgraded my video card tommorrow, Quake 2 benchmarks would on my to do list.

They both use the same engine. The original Unreal is just as CPU bound as UT,

I think not. UT has a new rendering engine which removes overdraw on the fly. The original Unreal doesn't do this. If you try Unreal (original) you'll see that it scales downward as you raise the resolution, just as any other 3D game does. UT stays the same regardless of resolution and regardless of video card (assuming it's half-decent).

Also, mixing the rendering libraries between the two games doesn't work at all. If the rendering engine was the same, it would work.

 

[BenSkywalker]

BFG-

"I've already seen the benches but I've also seen how it runs in real life. 60 FPS in Quake in real life sucks. 60 FPS in MDK2 in real life is more than playable."

60FPS for single player is what I look for, and a Celeron 300MHZ could deliver that with my gfx card at 1024x768 32bit UHQ. In fact, it wouldn't be any faster then my Duron is at the single player settings I use. But, the point I was trying to make is that MDK2 is not less strainful on systems.

"Not at all. All I know is how well it runs currently on my system. I also know that if I upgraded my video card tommorrow, Quake 2 benchmarks would on my to do list."

A GF2 GTS would likely give you 1600x1200 at playable FPS, a GeForce3 will do it without breaking a sweat. It may be important to you, but it should be an absolute given then any new board you buy, barring one of the budget offerings, are going to scream in Quake2.

"I think not."

It does, ask Epic. The last Unreal patch used the UT code, it is the same engine.

"If you try Unreal (original) you'll see that it scales downward as you raise the resolution, just as any other 3D game does."

I have tried it. Going from 320x240 all the way up to 1280x960 results in a whopping 12FPS drop, about the same that UT does(actualy, UT scales quite a bit more using Rev's bench). The only way to get a meaningful drop in FPS running the original Unreal is by enabling FSAA(those scores are rather old, pre Det3).

"Also, mixing the rendering libraries between the two games doesn't work at all. If the rendering engine was the same, it would work."

Not at all, I can't mix the renderers from different versions of UT, and that definately uses the same engine as itself I've actualy spent quite a bit of time with the rendering libraries for UT, it is the same engine as the original Unreal.